HAL Id: halshs-00448836https://halshs.archives-ouvertes.fr/halshs-00448836

Submitted on 20 Jan 2010

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Shona cl. 5 revisited: a case against ri- as a class 5nominal prefix

Michel Lafon

To cite this version:Michel Lafon. Shona cl. 5 revisited: a case against ri- as a class 5 nominal prefix. Zambezia, Jnal ofthe Univ. of Zimbabwe, 1994, 21 (1), pp.51-80. �halshs-00448836�

Zambezia (1994), XXI (i).

SHONA CLASS 5 REVISITED:A CASE AGAINST *RI- AS CLASS 5 NOMINAL PREFIX

MICHEL LAFON

French Institute for Research on Africa (IFRA)1

AbstractIn a critical reappraisal of the treatment of Shona noun classes, I challengethe hypothesized form of class 5 nominal prefix, *ri-, as it hardly everappears as such in the language. A comprehensive view of class 5 mergesclass 21 into class 5 and suppresses the need for a 'prosthetic' element toaccount for monosyllabic stems. Class 5 nominal prefix then includes allallomorphic realizations of the nominal prefix, from i- or ri- in monosyllabicstems, to voicing and zero as well as zi- with vowel<ommencing stems;augmentatives which exhibit voicing and zi-. The copulative construction,which calls for a prefixed i- in class 5, further suggests that i- is simply thelonger form of this class nominal prefix, and provides grounds for positing i-as a cover form.

* Indicates Proto-forms.' I could not have written this article but for the stimulus, advice and financial support I

received.My research took place during an initial two-year stay in Zimbabwe under the auspices

of the French Institute for Research on Alrica (IFRA).I would like to express my gratitude to the Directors of the Institute for allowing me to

devote my time purely to my research, and to the numerous Shona people whose kindnessand readiness to help encouraged me in my endeavours to understand and to use theirlanguage. Thanks are especially due to my friends, Mr. I. Katsiga from the Mount Darwin area,and Miss B. Manyuki from Gutu, both of whom patiently endured my questioning and provedvaluable collaborators.

Dr Chimhundu and his colleagues in the Department of African Languages and Literatureat the University of Zimbabwe kept doors of learning open to me, and Dr Chimhundu'sconcern for my work and his remarks on this article are duly appreciated.

Professor Fortune's MA seminar at the University of Zimbabwe in 1992, stimulated meto raise some of the issues dealt with here. His kind and detailed comments on an earlierversion gave me the opportunity to amend some of its more obvious defects, as well as adjustmy criticisms — my debt to him is great. Fortune's prominence in the domain of Shona studiesmakes him an obvious target for criticism. I would like to acknowledge his understandingattitude in this matter, which is further evidence of his keen dedication to improving ourknowledge of the Shona language.

I also appreciate the enlightened and enthusiastic comments of Dr Jefferies, a lecturerin linguistics and an expert on Shona.

A shortened version of this article was presented as part of a seminar series organizedby the University of Zimbabwe's Department of African Languages and Literature in August1993; 1 am grateful for the comments made on that occasion by the members of the Department,some of which are reflected here.

I stress that I take full responsibility for all mistakes, shortcomings and inaccurateinterpretations as may be encountered herein.

51

52 SHONA CLASS 5 REVISITED

WHEREAS THE VARIOUS realizations of Shona class 5 nominal prefix havebeen adequately described, some of the most extensively used Shonagrammars extend the concordial marker *ri- to cover the nominal prefix inthis class as well, since this is the hypothesized original form although ithardly ever appears as such in the language. This view appears to followFortune's descriptions in his prominent An Analytical Grammar of Shona2

and his Shona Grammatical Constructions,3 which largely expose theinfluence of Doke.4 Other works, such as Dale's Shona Companion andShona Mini Companion5 and various textbooks6 appear to have drawnheavily on Fortune.

1 suggest that Fortune's representation is misleading, and that class 5nominal prefix would be more adequately symbolized as a cluster ofallomorphic realizations or, as Carter calls it, 'a prefix scatter'7 coveringthe whole range of class 5 occurrences. This leads to a reassessment ofthis class so as to include class 21, and to a re-examination of thecopulative construction. The copulative construction neatly falls in withthe comparative hypothesis which suggests that i- is one among severalallomorphs. In considering this hypothesis, classes 9 and 10, which showrelated phenomena, will also be briefly discussed.

1 PRELIMINARIES: THE CLASS SYSTEM

I assume, in the following, that the reader is familiar with the basic featuresof Bantu languages and the main trends in the descriptions of Shona todate.

In line with Doke's and Fortune's views, the various forms of Shonaclass markers required by grammatical elements in agreement aredifferentiated to the extent that each form showing the slightest hint ofspecificity is accorded a category of its own. More than seven forms havetherefore been established, each matching a different category namely:noun, adjective, enumerative, quantitative, selectors, demonstrative andpronoun markers.8

2 G. Fortune, An Analytical Grammar of Shona (London, Longman, 1955).3 G. Fortune, Shona Grammatical Constructions,, Vol. 1 (Harare, Mercury Press, 3rd edn.,

1985) and Shona Grammatical Constructions , Vol. 2 (Harare, Mercury Press, 3rd edn., 1984).4 C. Doke, A Comparative Study in Shona Phonetics (Johannesburg, University of the

Wltwatersrand, 1931).5 D. Dale, Shona Companion: A Practical Guide to Zimbabwe's Most Widely Spoken

Language (Gweru, Mambo Press and Literature Bureau, 2nd edn., 1972) and Shona MiniCompanion: A Guide for Beginners (Gweru, Mambo Press, 1984).

6 V level Shona Grammar (Gweru, Mambo Press, 1970 and 1981).7 H. Carter, 'Introduction', in H. W. Chitepo, Soko Risina Musoro (London and Cape Town

Oxford University Press, 1958), 1-14.8 Fortune, Shona Grammatical Constructions , Vol. 1, 29-31, 116 and 121.

M. LAFON 53

This mode of organizing data, although very accurate and leavinglittle if anything unmentioned, fails to capture the inherent identity andunity of concordial agreements which Shona, in common with other Bantulanguages, very evidently displays. This is the result of an over-emphasison minor phonetic differences, often reducible to largely predictableadaptations, depending on whether the concords cling to a vowel or aconsonant, as well as to dialectal and individual variations.9

Indeed, if provision is made for nominal prefixes in classes 9, 10 and21, Shona class markers readily fall into two sets of forms:

i) a nominal set (termed 'nominal prefixes'), pertaining to nounsand adjectives, having only one allomorph specific to a fewnouns; and

ii) an agreement or concordial set (referred to as 'class concord'),for all other items, with the exception of class 1 which resortsto a range of different markers dependent on specificgrammatical units.10

Table I provides the basic forms of the two sets of class markers in theZezuru variety of Shona, with consonant-commencing stems.11

The class 5 nominal slot is left purposely empty, and will be dealt within due course. Class 9 and class 10 nominal prefixes, grouped here on thebasis of their intrinsic similarity, are given as a scatter of realizations:

i) nouns only: class 10:dzi- as a vestige for the pair of over-repeated instances, namelyimba/dzimba, 'house'/'houses', (ruimbo)/dzimbo, 'song'/'songs',as well as the few occasional cases of secondary use.12

ii) nouns and adjectives: class 9 and class 10:N- for nasalization, which is discussed later.

9 Another explanation may be traced to the comparative perspective favoured in theseaccounts whereby any category identified in one southern Bantu language is thought tosupport the itemization of corresponding elements in related languages, with little regardfor the specificity of each case.

10 This is in line with Guthrie's classical distinction between two main categories, verbalsand nominals. Nominate are further grouped into nouns, labelled 'independent nominals' andadjectives, labelled 'dependent nominals'. See M. Guthrie, Common Bantu (London, OxfordUniversity Press, 4 vols., 1967), iv, 221-8, and i, 13-14.

11 I follow standard Shona spelling, as in use in 1992.12 Fortune's description of the nasal in the two nouns as an instance of the class 9 prefix

does not seem valid as against its being lexical (Shona Grammatical Constructions, Vol. 1,66). In the case of dzimbo, the lexical view is supported by related words such as ruimbo'song', and -imba, 'to sing'. The interpretation of imba/dzimba follows by analogy —although there is no apparently related stem in Shona to substantiate this claim.

Therefore, Fortune's derived core form of class 10 nominal prefix, formed by theaddition of class 10 prefix \dzi-\ or 10-1 to the class 9 prefix I/V-I, seems over-complicated. Thepositing of \dzi-\ is based only on the two nouns above; all other instances where \dzi-\ doesnot surface are analyzed as 10- yv-l. This analysis serves as an example of conclusions beingdrawn from languages other than the one under scrutiny.

54 SHONA CLASS 5 REVISITED

0- which, according to Mkanganwi13 and Ngunga14 is used forwords, mostly borrowed nouns, where there is no change atall./- for monosyllabic stems, as will be suggested in the course of

this article.

Table I

CLASS MARKERS WITH CONSONANT-COMMENCING STEMS IN ZEZURUClass

123456789

10111213141516171821

Nominal Prefix*

mu-va-mu-mi-

ma-chi-zvi-N-; 0-; i-N-; 0-; i-; (dzi)ru-ka-tu-hu-; u-*ku-pa-ku-mu-zi-

Concords: Verb, -no- tense

a- / -mu- / u-'va-u-i -

ri-a-chi-zvi-i-dzi-ru-ka-tu-u-ku-pa-ku-mu-ri-

t Class 19 (sw-), almost entirely restricted to the Mhari dialect of Karanga,15 and class20, which goes unregistered in most accounts of Shona noun classes, are disregarded here.

* a- for subject prefix, -mu- for object infix, and u- for the remnant (in some dialects, it-is used in relative forms).

* In Zezuru, class 14 nominal prefix u- is restricted to disyllabic stems.

Some linguists have claimed an even greater degree of unity betweenthose two sets on the grounds that:

13 K. G. Mkanganwi, 'An outline of the morphology of substantives in Ndau with a preliminarynote on Ndau phonology' (London, University of London, M.Phil thesis, 1973).

14 A. Ngunga, 'A comparative study of Shona and Yao noun classes' (Harare, University ofZimbabwe, BA dissertation, 1987), 34.

15 Fortune, An Analytical Grammar of Shona, 103.

M. LAFON 55

i) not all classes show differences; if we consider class 9 andclass 10 main allomorphs (class 5 still excluded), only 6 classes,namely 1, 3, 4, 6, 9, and 10, out of the 18 presented here showany differences,

ii) all classes whose nominal prefix and class concord differ areendowed with a nasal in their nominal prefix and,

iii) with the exception of locative class 18 where the nasal isretained throughout, the concordial set appears to be thenominal one minus the nasal.16

Accordingly, these linguists posit Bantu languages as having onesingle set of class markers: viz. nominal ones, to which phonologicalrules are applied in certain instances, leading to the loss of the nasal inthe concordial set, locative class 18 being disregarded altogether.17

Notwithstanding the likelihood of this analysis in a diachronicperspective, Heny, within a generative framework, has dispensed withits supposed explanatory power as far as present-day Shona is concerned,since it would necessitate framing two rather ad hoc phonological rules,accounting for the non-syllabic allomorphs of classes 5, 9 and 10.Stating that 'none of them [operate] synchronically beyond this particularphenomenon', Heny convincingly argues the case for positing twodifferent, albeit related, sets of markers which he terms the 'two-prefixsolution', on the grounds of the syntactic and semantic specificity ofnouns and adjectives. He classically analyses adjectives as a sub-class ofnouns, as opposed to concords.18

I will, likewise, avoid the extreme position of accepting a single set ofclass markers, and retain two different series, related, but not to theextent that the relationship can easily be subsumed into a single series ofphonologically linked forms.

16 The odd behaviour of class 18 may tentatively be linked to the specificity of locativeclasses which operate, in most cases, at phrase level (A. Jefferies, personal communication).

17 Fortune's statement that 'the affixes should be regarded... as being the same morphemeas the noun prefix' is probably based on this rationale, although he does not elaborate.Fortune, Shona Grammatical Constructions, Vol. 1, 15.

18 F. Heny, 'Bantu lexical classes and semantic universals' (with some remarks on how notto write phonological rules), Studies in African Linguistics, (1972), iii, 222 and 231. Henyallocates the nominal prefix to 'deep structure', whereas class concords spread from the nounonto other agreeing elements (p., 256). Heny's paper has fuelled a stimulating controversy, inwhich Givon (T. Givon, 'On cost accounting in lexical structure: A reply to Frank Heny', Studiesin African Linguistics (1972 ), iii, 427-31) contests the grouping of nouns and adjectives simplyon the basis of concords when other criteria obviously separate them. Bennet (P. R. Bennet,'Heny vs. Givon: Pardon, may I cut In?', Studies in African Linguistics (1973), iv, 219-22) servesas a reminder that the setting of grammatical categories should arise from syntactic andsemantic behaviour and not from incidental morphological features. However, as this articledeals only with morphological Issues, Bennet's argument can be disregarded.

5 6 SHONA CLASS 5 REVISITED

2 TRADITIONAL VIEW OF CLASS 5 NOMINAL PREFIX: *R1-

For the sake of providing grammatical outlines, it is deemed convenientto ensure a representation of each and every class marker suitable as anentry to a table. In the case of Shona class 5 nominal prefix, followingFortune's influential analysis, this practice has resulted in resorting to aform identical to the class concord; that is, *ri-, in spite of the obviousdiscrepancy between the realizations and the chosen symbol.19

Fortune's argument is explicit in An Analytical Grammar of Shonawhen he states:

The prefix of class 5 never appears as such in the form ri- but its original presencein this form can be deduced from the forms of this prefix in other languages, Nguni,Tsonga and Sotho, from the forms taken by the concords of this class, e.g. theenumerative concord, and from the nature of the sounds which have resulted fromthe fusion of this prefix with the initial phones of the stem.20

In his more recent Shona Grammatical Constructions Fortunepursues the same line of argument, although he dismisses the use ofcomparative evidence, construing *n-, 'because this is suggested by theprefixes of other substantives of class 5, viz, by other forms of the classaffix'.21 Examples he then cites include -mwe, the enumerative, and -no,the close proximity selector.22

It seems that Fortune, as is the case with Doke,23 assumes theenumerative stands for any element exhibiting ri- as its class 5 concord,and that 'the reference to the enumerative concord has no nominalovertone'.24

The similarity of the two series of class markers has already beendismissed. I feel, however, that since the arguments postulated havefostered the widespread misconception of *ri- as the class 5 nominalprefix echoed in the well-known rhyme which strings together nominalprefixes, running mu-/va-, mu-lmi-, ri-/ma-, chi-lzvi-, and so on, thesearguments, popular and likely though they may be, especially in relationto the enumerative stems, need to be re-examined.

19 Hazel Carter makes no attempt t o construe a single symbol for this prefix, s imply statingthat it has disappeared, whereas the change in stem-initial betrays its ex i s t ence (H. Carter andG. P. Kahari, Shona Language Course (London, School of Oriental and African Studies , 1972), 7.Other scholars posit zero — Mkanganwi, "An outline of the morpho logy of substant ives ..'.',I13-14; and Ngunga, 'A comparative s tudy of Shona and Yao noun c lasses ' , 23.

20 Fortune, An Analytical Grammar of Shona, 53.21 Fortune, Shona Grammatical Constructions, Vol. 1, 32 and 50.22 Ibid., 116.23 Doke, A Comparative Study in Shona Phonetics, 125.24 Fortune, personal communication.

M. LAFON 57

2.1 Internal evidence: The enumeratlveLet us examine Fortune's evidence in support of *ri- as the class 5 nominalprefix viz; the behaviour of 'enumerative' stems.

Fortune identifies three so-called 'enumerative stems', -mwi, 'same', -mwe, 'other', 'a certain', 'some', -i ? 'of what sort?'25 As -i seems to be solittle used in present-day Harare speech as to prescribe any clearelaboration,26 we shall contemplate only the two segmentally identical,monosyllabic stems -mwe and -mwe, which can be treated together sincetheir concords are alike in all cases.

These concords, as listed in Fortune's An Analytical Grammar ofShona,21 offer an alternative for all classes the nominal prefix of whichcontains a nasal consonant: 'there are two types of concords, "strong"(containing a nasal consonant), and "weak", (not containing a nasalconsonant)'.28 The distribution of these forms follows dialectal (andprobably also idiolectal) lines. The 'strong' series corresponds with thenominal prefix, and the 'weak' series with the class concord. Hence, if wedisregard class 5 and class 10, this alternative amounts to a choice betweenthe two main series of markers, and these stems can be said to behaveeither nominally or concordially, as shown by Table II below.29

This leaves out class 5 which, even for those dialects which requirethe nominal or 'strong' forms, consistently displays the concordial series.The same applies, to a lesser extent, to class 9 and class 10, furthercomplicating the issue.

If we relate the enumerative markers to the two basic sets, the'unexpected' presence of the class 5 concord in place of the nominal prefixcould be explained as an extension of the concordial agreement when, forphonological reasons, the nominal prefix fails to operate. The form whichwould have resulted from implementing the feature exhibited bymonosyllabic stems in class 5, which might be hypothesized as *imwe,with 'epenthetic' [i] (discussed in greater detail later), is not documented.Furthermore, since the nature of the initial consonant of the stem precludes

25 Fortune,Shona Grammatical Constructions, Vol. 1, 111.26 The only consistent context I was able to elicit is In enquiring the sex of a person,

especially a new-born baby: mwana muytf 'which (sort of) child?' In all other instances,either the interrogative -pi or the encliticization of -/, was preferred.

27 In his Shona Grammatical Constructions, Vol. 1 only the 'strong' forms are given.28 Fortune, An Analytical Grammar of Shona, 165-6. Note that class 18 stands out again and

belies this suggested trend as it maintains the nasal even in the class concord.29 This view may b e supported by the very controversial concord in zi-, suggested here as

an alternative to ri- in so-called class 21. Out of a brief survey of about ten people fromdifferent areas, zimwe would be reluctantly tolerated by only two, while one would cite It asrevealing a non-native speaker.

58 SHONA CLASS 5 REVISITED

voicing, dialects using the nominal or 'strong' set for those items wouldhave had to resort to use of the concordial or 'weak' set to ensure a properand overt mark in that class, either as a result of internal analogy orthrough influence from neighbouring varieties.

Table II

CHART SHOWING -MWE CONCORDS IN ZEZURU, KARANGA AND MANYIKA

Class Nominal Prefix Class Concord Class Nominal Prefix Class Concord'strong forms' 'weak forms' 'strong forms' 'weak forms'Zezuru & Manyika Zezuru & Manyika30

Karanga Karanga

123456789

10

mumwevamwemumwemimwerimwemamwechimwezvimweimwedzimwe

umwevamweumweimwerimweamwechimwezvimweimwedzimwe

111213141516171821

rumwetumwekamwehumwekumwepamwekumwemumwerimwe(zimwe)

rumwetumwekamweumwekumwepamwekumwemumwerimwe .

For class 9 and class 10 the case is more straightforward since: i) thebehaviour of the stem in class 9 is no different from that of monosyllabicstems, and ii) the class 10 form may be assimilated to the few instanceswhere a prefix dzi- appears in nouns. These forms can therefore beconsidered as instances of nominal agreements.

The two enumerative stems would then conclusively exhibit eithernominal or concordial agreement, phonology permitting.

From this evidence, should ri-, i- and dzi- be considered genuinenominal prefixes for classes 5, 9 and 10, they could only be posited asalternative forms of the nominal prefixes. This does not, however, provideany grounds for the theory that they are anything other than lexicallydistributed allomorphs.

30 This division, based on Doke's pioneer survey, should not be taken as clear-cut: it iskept more for reasons of convenience than accuracy. A certain amount of overlap has beenpointed out. For instance, not all Manyika-dialect speakers use the 'weak' forms, Mahungwespeakers, located on the border with Zezuru-speaking areas, have the 'strong' ones, and so on.

M. LAFON 59

2.2 External evidence: Comparison with Bantu languagesEven though the comparatist argument was not maintained by Fortune, itmay be useful to bear in mind that data taken from related languagesshould not be decisive in framing the grammar of a specific language.Therefore, neither the similarity between class 5 nominal prefix andclass concord exhibited, to some extent, by other Southern Bantulanguages, nor the forms of the nominal prefix in those languages, give anydependable grounds for making assumptions about Shona if clear examplesare not found in that language itself.

Moreover, major comparative works on Bantu languages posit twodifferent reconstructions for class 5 inflected elements: the nominal andthe concordial. Guthrie construes class 5 nominal prefix as *yi or yi,class 5 concord as *di or di, and Meeussen as *i and *di respectively.31

Having strengthened our argument for positing the two series ofmarkers, we can now address the specific issue of class 5 nominal prefix.

3 CLASS 5 IN PRIMARY ASSIGNMENT

A review of Shona class 5 nominal prefix allomorphs is needed in orderto assess whether or not the commonly accepted representation isaccurate.

Class 5 nominal prefix is usually described as consisting of twoallomorphic realizations of the hypothesized cover-form *ri-, accordingto whether the stem is consonant- or vowel-commencing: voicing and zi-respectively.32 Augmentatives with prefix zi; in spite of commandingclass 5 concord, are set aside in a specific class numbered 21.

The extensive view of class 5 nominal prefix that I advocate hereincludes augmentatives, which use the very same prefix allomorphs, butin a broader context, as well as the device required by monosyllabicstems, /- (rarely n-). I shall review them all and consider what 'coverform' they suggest for the nominal prefix.

In line with Fortune, I differentiate prefixes in primary function,which are normal, from those in secondary function, which areoccasional, opposing reference meaning to a connotative based on it.33

Within class 5 primary usage we need to establish a distinctionbased on stem-length, separating monosyllabic stems from disyllabic ones.

31 Guthrie, Common Bantu, Vol. IV, 221-8; and A. E. Meeussen, 'Bantu grammaticalreconstructions', in J. Doneux et al. (eds.), Africana Linguisticana (1967), Hi, 97.

3 2 No explanation whatsoever is offered for derivation of zi- from *ri- which seems anunlikely phonetic result. Fortune, Shona Grammatical Constructions, Vol. 1. 50.

3 3 Fortune, An Analytical Grammar of Shona, 54.

gO SHONA CLASS 5 REVISITED

Whereas voicing has been extensively described, little emphasis seemsto have been devoted so far to other class 5 nominal prefix allomorphs,such as i- for monosyllabic stems, and zi- for vowel-commencing ones.

3.1 Monosyllabic stems: i- (0-, ri) as a nominal prefixMost Shona dialects do not admit monosyllabic words, Karanga being oneexception.34 Therefore, when in isolation, otherwise monosyllabic class 5nouns are supplemented with an initial vowel I i I, allowing them to beuttered. This vowel is said to embody the 'stabilizer', another instance ofwhich is found in monosyllabic imperatives:

idyal, 'eat!' < -dya, 'to eat', but tora\ 'take!', < -tora, 'to take'.35

In nominal phrases, this stabilizer becomes redundant, and can bediscarded, as will be shown below.

i- has been considered as a separate element termed 'penultimate'36

or 'prosthetic'37 whereas Stevick more aptly refers to it as a 'dummy prefixfor nouns lacking prefix'.38

1 contend that i- is yet another realization of a class 5 nominal prefix,and that creating a special one-element category to accommodate it ispointless. Instead, we have an instance of a nominal prefix functioning as astabilizer. To substantiate my claim, the absence of i- in certain phrasesmust be addressed, as this behaviour is atypical of a true prefix in Shonaand would explain the reluctance of grammarians to assimilate it to aprefix.

I will examine the conditioning of this vowel in both words in isolation,and in phrases.

3.1.1 Words in isolationThe i- vowel, present with monosyllabic class 5 stems, is deleted in theplural class 6 marked by prefix ma-.

izwi, *word' mazwi ibwe, 'stone' mabweibvi, 'knee' mabvi ivhu, 'soil' mavhu

A similar phenomenon occurs in class 9 and class 1039 which show nodifference in their nominal prefix:

34 Ibid., 74 a n d For tune , Shona Grammatical Constructions, Vol. 1, 50.35 < implies derivation from t h e form t o t h e right of it.36 Doke, A Comparative Study in Shona Phonetics, 125 and For tune , Shona Grammatical

Constructions, Vol. 1, 50.37 Fortune, An Analytical Grammar of Shona, 74.38 E. W. Stevick a n d K. Machiwana, Manyika StepJryStep (Cleveland, Transvaa l , T h e Central

Mission Press , 1960), 107.39 I include class 9 and class 10 as this gender presents the same situation as that of class

5. In order for it to be valid, any explanation that dispenses with having an ad hoc element forclass 5 i- should also accommodate classes 9 and 10.

M. LAFON 61

classes 9 and 10, imbwa (-), 'dog'; inda (-), 'louse';class 9, imba, 'house'.40

The only monosyllabic adjectival stem recorded is consistent withthis distribution, its initial also being affected by voicing in class 5:

•tsva, 'new': class 5 idzva, class 9 and class 10 itsva, class 6 matsva,class 4 mitsva, and so on.For class 5, an alternative form ri- is encountered in some dialects

in what is evocative of a lexical distribution. The following examples arebriefly mentioned by Doke as instances of the rolled consonant in Tavaraand Ndau:

ribwe for ibwe, 'stone'; ribvi for ibvi, 'knee'but Mi, 'truth' (*ridi).*x

Hannan records a few odd cases such as rifa/mafa, 'inheritance', inKaranga,42 and this is confirmed by Fortune.43

As ri- can hardly be referred to as a penultimate or prostheticelement, this distribution suggests that /-, ri-, perhaps zi-, as well as theKaranga variant 0- constitute a set of class 5 nominal prefix allomorphsrequired by monosyllabic stems and distributed along dialectal lines.Therefore, if one accepts this first step, i- and 0- can also be posited, byanalogy, as class 9 and class 10 nominal prefix allomorphs.

3.1.2 Nominal phrases with monosyllabic stemsIn Shona, nouns in phrases may be preceded by a 'latent'44 or pre-initialvowel, according to the nature of the phrase.

Although it does not ultimately interfere with the issue in question,the data necessitates differentiation of noun phrases in which the latentvowel is implied from those in which it does not appear.

40 (-) indicates that no change occurs in the formation of the plural.41 Doke, A Comparative Study in Shona Phonetics, 74.42 M. Hannan, Standard Shona Dictionary (Harare, College Press and Literature Bureau,

1984).43 Fortune, Shona Grammatical Constructions, Vol. 1, 50. These examples were readily

accepted by participants at the University of Zimbabwe seminar, although they favouredmarifa as the plural.

Another class 5 nominal prefix allomorph for monosyllabic stems may well be zi-, as Inzihwe (mahwe), 'stone', mentioned by Buck with no augmentative shade [Buck, Rev.], ADictionary with Notes on the Grammar of the Mashona Language Commonly Called Chiswina(Penhalonga, St. Augustines, SPCK, 1911), vl, 206. The words ziko, 'fireplace', and zifa, 'sourmilk', given by Hannan, may be indicative of older or dialectal forms, or borrowings fromNgunI as in the case of ziko at least.

The plural formation in which the first syllable of the word tends to remain unchanged{mazifd), suggests the progressive l e g a l i z a t i o n of the prefix.

44 Fortune, Shona Grammatical Constructions, Vol. 1, 53.

6 2 SHONA CLASS 5 REVISITED

3.1.2.1 Nominal phrases with influence of latent vowel: -a, naWhen preceded by [a] ending elements, such as the 'associative' -a (aconstruction equated to the possessive) or na, meaning 'and', 'with', or'by', nouns entail the coalescence of this vowel [a] with the latent vowel,which is characteristic of each noun class, the quality of which may bededuced from the output.

With syllabic prefixes, the latent vowel is in harmony with the vowelof the nominal prefix, resulting accordingly in either one of the followingvowels [a, o, e]:45

basa ravana, 'the work of the children'; class 2 latent vowel [a]:\ra+a+va-...\nomukadzi, 'with the woman': class 1 latent vowel [u]: \na+u+mu-...Ipamberi pomugwagwa, 'further down the street': class 3 latent vowel[u]: \pa+u+mu-...\mukati mechidzidzo, 'in the midst of the lesson': class 7 latent vowel[I]: \ma+i+chi-...\ane zvitoro, 'he has shops': class 8 latent vowel [i]: I na+i+zvi-... I.

The quality of classes 5, 9 and 10 latent vowel is [i], as can beinferred from the instances where [e] surfaces;46

ane basa, class 5 'he has work'; nguva dzehondo, class 9 'the times ofthe war'; nehama dzake, class 10 'with his relatives'.

Monosyllabic stems in the above phrases behave in a similar mannerwhile allowing for the deletion of the prefix, as signalled by Mkanganwifor Ndau47 — note that this variation seems devoid of meaning:

class 5: neibwe/nebwe, 'with the stone'; mwana weivhu/wevhu, 'sonof the soil';class 9: neimba/nemba, 'and the house'; musoro weimbwa/wembwa,'the head of the dog'; mukati meimba/memba, 'inside the house';class 10 ane inda/nda dzakawanda, 'he has many lice'.

We turn now to phrases where this vowel does not occur.

45 Classical Zezuru offers a wide and systematic range of variation: [a*a] > [a ] , [a+u] > [o ] ,[a+1] > [e ] . Some dialects have neutralized this variation resulting in either [e] or [o ] for allexcept proper nouns . I purposely o b s e r v e the broader s c o p e of realizations.

46 I i I is not sugges ted as being the form of the nominal prefix: the data mere ly prove thatthe nominal prefix vowel is of the s a m e quality as the latent vowel .

47 Mkanganwi, 'An outline of the morphology of substant ives . . . ' , 100.

M. LAFON 63

3.1.2.2 Nominal phrases with no influence of latent vowel: locativeconstructionIn locative construction, the latent vowel does not arise and the vowel ofthe formative surfaces as it is ([a], [u]):

pamugwagwa, 'on the road'; pachikoro, 'in the school';mumugwagwa, 'on the road'; kuchikoro, 'towards the school'.

Here again, monosyllabic stems maintain alternative forms, accordingto the presence or absence of the vowel which I claim is prefixal:

paibvi/pabvi, 'on the knee'; muibvi/muboi, 'in the knee';paivhu/pavhu, 'on the soil'; muivhu/muvhu, 'in the soil';paimba/pamba, 'at the house'; muimba/mumba, 'inside the house'.

Coalescence between the vowel of the formative and that of thenominal prefix seems impermissible within the language (*pebvi, *pemba,and so on).

Within our hypothesis of i- as an allomorph of classes 5, 9 and 10 wewould account for alternative forms by including an optionalphonological rule which permits deletion of this V-shaped allomorphafter a vowel:

CV-iC... > CV-iC... or CV-C...

The rule would simply state that the I i I allomorph of class 5 andclass 9 nominal prefixes may be omitted in monosyllabic stems inphrases after a vowel. Since I i I would be the only vocalic nominal prefix,the rule could be generalized and restated as optional deletion of thenominal prefix vocalic allomorph of monosyllabic stems in phrases.

This optional phonological rule, sensitive to lexical category, seemsplausible if one remembers that [i] is a weak vowel which is deleted inmany contexts. Concord markers of Ci- shape, which become C- withvowel-commencing items, viz, subject markers before I-a-1 (past) or l-o-l(exclusive), or the common realization of masikati, [maskati], 'goodafternoon'/'afternoon', are but a few examples of this deletion.

An even greater degree of generalization encompassing both situationsmight then make provision for the deletion, optional or compulsory, of[i] belonging to a class marker in a vocalic sequence.

In traditional descriptions, short forms (monosyllabic words) wereseen as basic, and what appeared to be the supplementation of a vowel tothese short forms had to be explained along these lines — hence the needto resort to penultimate or prosthetic features.48 Conversely, positing the

48 This view is supported by other examples in the language.

64 SHONA CLASS 5 REVISITED

longer forms as basic, we need not cite specific elements for monosyllabicstems but only consider cases for deletion of elements; a process whichseems at least as natural as supplementation. Consequently, this viewobviates the need for specific items catering for monosyllabic stems, andenables a more general treatment of noun stems.

Assuming we now agree to consider <- surfacing in monosyllabic stemsas a nominal prefix allomorph of classes 5, 9 and 10 rather than as aseparate element, I will now turn my attention to class 5 disyllabic stems.

3.2 Disyllabic stemsIn disyllabic stems, class 5 nominal prefix is divided according to thephonetic nature of the stem-initial, which can be a vowel or a consonant.

3.2.1 Vowel<ommencing stems: z(i)-The classification of z(i)- as a class 5 nominal prefix allomorph withvowel-commencing stems is well-established. Fortune, for example,mentions it as one realization of ri-, listing a handful of vowel-commencing stems where the status of zi- as a prefix (and not the stem-initial) proves against plural (class 6) or other related forms.49

class 5 (singular) class 6 (plural)zano, 'plan' mano or mazanoziso, 'eye' meso or mazisozino, 'tooth' meno or mazino

zuro, class 5 'yesterday' mauro, class 6 'evening'zero, class 5 'age' muero, mwero, class 3 'amount'zambuko, 'ford' -yambuka, 'to ford'

The alternative plural formation which retains zi-, and is the formmost frequently used in urban Shona, provides a fair indication thatthis allomorph is probably in the process of losing its grammatical(prefixal) value and amalgamating within a restructured stem. Thisview is confirmed by the copulative construction as well as CV- prefixallomorphs in monosyllabic stems: examples of the latter have alreadybeen supplied for class 5.50

49 Fortune, An Analytical Grammar of Shona, 73 and Shona Grammatical Constructions, Vol .1,50.

50 See footnote 43.

M. LAFON 65

3.2.2 Consonant<ommencing stems: The voicing processMost nominals when entered in classes 5, 9 and 10 exhibit a change intheir initial consonant. Resulting from Doke's pioneer study, the class 5alteration is known as 'vocalization', over which Fortune's term 'voicing'is now often preferred. Classes 9 and 10 can be subsumed under'nasalization', referring to the basic form of the class prefix, even thougha nasal is not necessarily present in all cases.51

Voicing of the initial also does not appear to be a necessary conditionfor inserting nominals in class 5, as nominals may register in that classbearing no overt mark, as will be illustrated below.

The behaviour of the stem-initial is all the more obvious with wordswhose initials appear in different contexts, a definition which is ideallysuited to adjectives by nature and, to a lesser extent, nouns belonging togender 5 and 6,10 and 11, or de-verbals. The canonical shape of the stem-initial in cases of change can, therefore, best be captured in an inter-vocalic context. I shall review all consonants of the language, dividingthem according to their behaviour when functioning as stem-initials toclass 5 nouns.

3.2.2.1 Initials exemplifying changesThe main regular changes involving class 5 are given and exemplified inTable III below.52 I have supplied the corresponding class 9 and class 10forms for comparison (this gender modifies a number of other consonantswhich we will disregard, as they are not the subject of this discussion).The changes are fairly widespread, although one has to take cognizanceof dialectal and lexical variations.

51 Pongweni, among others, offers a study of the changes experienced by class 9 anqclass 10 de-verbal nouns and their phonetic representation in his Studies in Shona PhoneticsAn Analytical Review (Harare, University of Zimbabwe Publications. 1990). 44-53.

52 This presentation is heavily indebted to Doke's and Fortune's thorough accountswhere more examples can be found (Dolce, A Comparative Study in Shona Phonetics. 125-3)and Fortune, An Analytical Grammar of Shona, 72-9). Many authors have since attempted todraw a more synthesized and systematized view of the phonological processes involved. |acknowledge their endeavours but maintain a less elaborate description in this article.

66 SHONA CLASS 5 REVISITED

Table III

TABLE OF CHANGES UNDERGONE BY INITIAL CONSONANTS

Inter-vocalic

Classes 6, 2, etc.

Ptkchtsvsv(Pt

Voicing

Class 5

bdgJdzvdzv/sv53

bv

Nasalization

Classes 9 and 10

mhnhhchtsvsvPf)54

Adjectives whose stem-initial appears in the three contexts specifiedin Table III, provide sufficient data to exemplify all cases.55

-pamhi, 'large': class 3 muti mupamhi, 'a large tree'; class 5: gumbobamhi, 'a large leg'; class 9 tafura mhamhi, 'a large table';4ete, 'thin': class 1 munhu mutete, 'a thin person'; class 5 bepa dete,'a thin (piece of) paper'; class 9 nyama nhete, 'a thin (piece of) meat';•hum, 'big': class 1 munhu mukuru, 'a big person'; class 5 bere guru,'a big hyena'; class 9 shiri hum, 'a big bird';•chena, 'white': class 4 miti michena, 'white trees'; class 5 bere jena,'a white hyena'; class 9 shiri chena, 'a white bird';-tsvuku, 'red': class 3 muti mutsvuku, 'a red tree'; class 5 bangadzvuku, 'a red knife'; class 9 tafura tsvuku, 'a red table';-svinu, 'healthy', 'attractive'; class 1 musikana musvinu, 'an attractivegirl'; class 5 bere dzvinu/svinu, 'a healthy hyena'; class 10 nzousvinu, 'healthy elephants';-phimbu, 'grey': class 7 chigaro chipfumbu, 'a grey chair'; class 5bere bvumbu, 'a grey hyena'; class 9 tafura phimbu, 'a grey table'.

3.2.2.2 Non-affected initialsAs previously stated, some words are not affected by the voicing process:their stem-initial includes not only voiced consonants, which by definitioncannot undergo further voicing, but others as well. Moreover, some words,

53 Optional change In class 5 for s o m e words.54 This series Is poorly illustrated: Doke, A Comparative Study in Shona Phonetics, 129,

gives only one example, -phimbu, while Fortune suggests another, bvupa/mapfupa, 'bone' in AnAnalytical Grammar of Shona, 73.

55 See Dale's table of adjectives in Shona Companion, 30-31.

M. LAFON 67

mostly borrowed, display in class 5 in initial position a consonant pertainingto the voiceless set, thereby contravening the hypothesized conditioning.

a 'Unvoiceable' initials: Voicing is understandably excluded withvoiced initials phonically identical to the items of the voiced set, to whichmay be added the plosives [bh] and [dh], and miscellaneous other featuressuch as nasals, rolled consonants, and semi-vowels.

Adjectival stems again provide the most conspicuous illustration of'unvoiceable' initials; other instances include nouns in classes 5 and 6.

a (i) Voiced initials: Include:• items identical to those of the voiced set: [d, b, g, j , dzv, bv] and

[bh, dh]. Examples:-diki, 'small': class 1 munhu mudiki, 'a small person'; class 5 berediki, 'a small hyena'; class 9 nzou diki, 'a small elephant';basa/mabasa, 'work'; gumbeze/magumbeze, 'blanket';jasi/majasi, 'coat'; dzviti/madzviti, 'Ndebele person';bvute/mabvute, 'shade'; bhuku/mabhuku, 'book';dhorobha/madhorobha, 'town'.

• other voiced consonants: [v, vh, z, zv, zh, h, dz], some examples ofwhich are:vanga/mavanga, 'scar'; vhere/mavhere, 'pubic hair'; zizi/mazizi,'owl'; zvere/mazvere, 'maize plant';-zhinji, 'much', 'many': class 2 vanhu vazhinji, 'many people';class 5 basa zhinji, 'a lot of work'; class 9 mari zhinji, 'muchmoney';-hombe, 'big': class 1 musikana muhombe, 'a big girl';class 5 banga hombe, 'a big knife'; class 10 nzou hombe, 'bigelephants'; dziva/madziva, 'pool'.

a (ii) Miscellaneous consonants:• nasals [n, nh, m, ny, mb]. Examples include:

nwiwa/manwiwa, 'water-melon'; nhanga/manhanga, 'pumpkin'-nyoro, 'soft': class 3 muti munyoro, 'a soft tree'; class 5 dhakanyoro, 'soft mud'; class 9 nzungu nyoro, 'a soft ground-nut';-mbishi, 'unripe': class 6 maranjisi mambishi, 'unripe oranges';class 5 ranjisi mbishi, 'an unripe orange';class 9 nyama mbishi, 'uncooked meat'.

No example was found with [m] as initial, but it may be assumed tobelong in this category.

• rolled/r/:-rehi, 'tall', 'high', 'long': class 1 musikana murefu, 'a tall girl';class 5 gomo reni, 'a high mountain'; class 9 mhino ndeni or refu, 'along nose'.

68 SHONA CLASS 5 REVISITED

• semi-vowel /w/:windo/mawindo,56 'window'.

b Voiceless initials: Some voiceless initials are not affected by theprocess of voicing: this rarely includes consonants which are identical tothose which become voiced when entered into class 5, as well as a fewothers.

b 0) Consonants identical to the voiceless set [p, t, k, ch, sv, tsv, pf]:Hardly any nominals register in class 5 with an initial belonging to the

above set. The very few recorded cases include mostly borrowed wordsand it is interesting to note that, although their plural is in class 6, thesingular allows for variation in class allocation between classes 5 and 9.This proves conclusively that their initial is not really suited to class 5,and hints at the strength and effectiveness of the conditioning:

classes 5 or 9 purazi 'farm' — plural mapurazi — (from Afrikaans);classes 5 or 9 tikiti 'ticket' (from English);class 5 cheri 'cherry' (from English);classes 5 or 9 svondo 'week' or 'Sunday' (from English);57

kii is allocated to class 9 only in spite of the plural makii in class 6(from English).

Interestingly, the adjective -pfupi 'short' remains unchangedthroughout, irrespective of concord:58

class 1 munhu mupfupi, 'a short person';class 5 banga pfupi, 'a short knife'; class 10 nzou pfupi, 'short

elephants'.

b (ii) Other voiceless initials [f, s, sh, ts]: Voiceless consonantswhich are not part of the distribution [f, s, sh, ts] are unchangedalthough they do have voiced counterparts in the language:59

-shoma, 'few' or 'little': class 4 miti mishoma, 'few trees';class 5 basa shoma, 'little work'; class 10 nzou shoma, 'few elephants';fafitera/mafafitera, 'window'/'windows'; sadza/masadza, 'porridge';tsaga/matsaga, 'sack'/'sacks'.

56 Hwindo also occurs.57 Some speakers take advantage of the alternative class allocation to bypass the ambiguity,

and u s e svondo in class 5 for 'week' (svondo rino, 'this week1), and in class 9 for Sunday(svondo ino, this Sunday). However, usage among first-language speakers is inconsistent.

58 Its variant -fupi also remains unchanged; see Dale's Shona Companion, 30.59 The fact that some consonants are not voiced when they could be suggests the inaccuracy

of the term voicing to define the phonetic process involved perse. The data may here lead toa division of a number of the consonants into two sets: those amenable to change and thosewhich are not. More in-depth research is needed in this area.

M. LAFON 69

So far, class 5 nominal prefix, in primary function, may be summarizedas

• /-, ri- for monosyllabic stems• zi- for vowel-commencing stems• voicing for a limited set of voiceless consonants (with

lexical exceptions)• no change for other consonants

I now examine the evidence suggested by class 5 when in secondaryfunction.

4 THE AUGMENTATIVE: CLASS 5 IN SECONDARY ASSIGNMENT

As stated earlier, class 5 is both a primary and a secondary class. In thelatter capacity, it conveys an augmentative and often derogatory shadeof meaning.60

Augmentatives are obtained either by prefix substitution, orreplacement, or by the superimposition or addition of a prefix. Thisprocess involves two class 5 nominal prefix allomorphs, zi- and voicing(excluding, as one would expect, i- and ri-'), whose distribution assecondary prefixes mirrors their occurrences as primary prefixes, ifonly partially. The process is by no means clear-cut and allows forvariation as well as alternative forms in many given instances —augmentatives are a field quite naturally open to individual creativity,and this is often reflected in differing judgements by first-languagespeakers on the acceptability of a particular item.

It is clear that the setting up of a class, namely 21, distinct fromclass 5 only in its nominal prefix, to accommodate augmentatives withzi- prefix as favoured in the accounts of Shona class system underreview, was geared to facilitate the presentation of the system as a whole,not to mention the comparative aims. Construing class 21 also avoidsthe rather embarrassing feature of a class whose prefix can sequencetwice on the same item, whether in identical or in different allomorphs.61

However, it artificially splits the formation of augmentatives, separatingthose derived through voicing from those exhibiting prefix zi-. This rationalemight further lead to division of all classes which manifest both primaryand secondary use: namely classes 1, 2, 7, 8 and 11.

60 As I am concerned with morphological issues, I shall not dwell on the precise meaningof the augmentative examples, the general gloss of 'big' being sufficient.

61 Fortune, however, acknowledges such a possibility when class 10 with dzi- is posited as asecondary noun class, providing examples such as dzinzira, 'several paths' from nzira, 'path1,analyzed as dzi- class 10; O- c lass 10; N- class 9 in Shona Grammatical Constructions, Vol. 1, 66.

70 SHONA CLASS 5 REVISITED

I contend that the questionable gain in simplicity of presentation isnot sufficient justification for positing yet another class in an already richsystem, especially when this results in a confusing relationship betweenclass 5 and class 21.62

The alternative theory involving merging class 21 and class 5 has theadvantage of obviating the need for a class in which the adjectival prefixdiffers from the nominal one, thereby nurturing the artificial split betweenthe prefixes of adjectives and nouns.

My proposal can be viewed as a return to a more classical analysis,since previous writers used to foster a united account: Doke treated zi-as a member of a single *ri- class63 and pointed out Shona ri- (standingfor class 5) as an example of an augmentative prefix in Bantu, beforegranting it a class of its own in his later survey of Southern Bantulanguages.64

4.1 Prefix substitution: zi- and voicingLiterature on the subject, especially that of Fortune,65 upholds manycases of prefix substitution which sound unfamiliar, to say the least, tothe few Shona-speaking Harare youths interviewed. This would suggestthat these examples pertain either to a classical register (so-called 'deepShona'), an archaic stage of the language, or dialectal usage but thisdoes not detract from their validity as instances of the process at work.The present tendency in such cases seems to favour superimposition ofprefix.

The process of prefix substitution is as follows: the original nominalprefix is replaced by an allomorph of class 5 nominal prefix, in line withits distribution and linked, as has been shown, to the stem-initial.

4.1.1 Vowel<ommencing: z(i)-zana, 'big child' or 'naughty child'; compare — mwana, (stem *-and)66 class 1 'child';

62 For instance, Mkanganwi, 'An outline of the morphology of substantives . . . ' placesvowel-commencing s tems which have z(Q- as a nominal prefix in class 21 (zanda, 'egg' andzana, 'idea"), stating that they are the only words in which class 21 is primary.

63 Doke, A Comparative Study in Shona Phonetics, 126.64 C. Doke, Bantu Linguistic Terminology (London, New York, Toronto, Longman, 1935), 57

and C. Doke, The Southern Bantu Languages (London, Oxford University Press, IAI, 1954), 52.65 Fortune, An Analytical Grammar of Shona, 104.66 'compare' is used here to refer to another instance of the same stem, whence the

augmentative cannot easily be considered as derived. Herbert argues the same case as thatgiven here by 'compare', and goes even further when he says all connotatives obtainedthrough prefix substitution 'are not synchronically derived by rules from the normal nounclass forms'. See R. K. Herbert, 'Prefix restructuring: Lexical representation and the Bantunoun', Studies in African Linguistics (1977), Supplement 7, 106. However, this view s e e m s tobypass conclusions which could legitimately be reached by a non-native speaker.

M. LAFON 71

zoto, 'large fireplace'; compare — choto, (stem *-otd) class 7 'fire-place';zenga, 'large frying pan'; compare — rwenga, (stem *-engd)67 class 11'frying pan';zinhu, 'big thing'; compare — chinhu, (stem *-nhu) class 7 'thing'.

4.1.2 Consonanteommencing stems: When the stem-initial is aconsonant, depending on its phonetic nature, entering the noun into class5 entails the same possibilities as seen when class 5 functions as a primaryclass.

4.1.2.1 Initial belonging to the voiceable set: Voicing The initial whenbelonging to the voiceable set, may undergo voicing although this processis far from systematic:

gadzi, 'big woman' < mukadzi, class 1 'woman'; gomana, 'big boy' <mukomana, class 1 'boy';danda, 'big log' < mutanda, class 3 'log';bini, 'big handle' < mupini, class 3 'handle'; bino, 'big nose' < mhino,class 9 'nose';jembere, 'old hag' < chembere, class 9 'old woman'.Some constructions, although apparently phonologically permissible,

are belied by usage:*guyu < mukuyu, class 3 'fig tree'; *bembere < chipembere, class 7

'rhino'.

Some other changes have been recorded, especially the voicing andstrengthening of [s] into the affricate, [dz]:

dzikana, 'large girl' < musikana, class 1 'girl'; dzoro, 'big head' <musoro, class 3 'head'.

4.1.2.2 Unvoiceable initials: No change Uncharacteristically and withvarying degrees of acceptability by first-language speakers, a few nounswhose initials are not amenable to voicing may be lodged in class 5,experiencing no change of initial:

(T)68 rume, 'undesirable man' < murume, class 1 'man';(?) sikana, 'large girl' < musikana, class 1 'girl'.69

67 I w a s unable to verify t h e s e last t w o examples with any of my informants.68 (?) indicates a doubtful instance/feature.69 Fortune, An Analytical Grammar of Shona, 77. However, I recorded only dzikana with

reinforcement of the initial, quoted above .

72 SHONA CLASS 5 REVISITED

4.2 Superimposition of allomorph zi-The other device yielding augmentatives is superimposition of theallomorph zi-, which seems a more popular option than prefix substitutionand is more commonly used in everyday speech. Zi- is added to the frontof nouns in several different classes, class 5 included. This gives rise to asequence of (at least) two class prefixes.

The addition of zi- is most conspicuous with syllabic prefixes, butaffects non-syllabic ones as well.

4.2.1 With syllabic prefixesThe superimposition of the allomorph occurs consistently with class 1and class 3 nouns, and more rarely with class 11:

zimunhu, 'big person' < munhu, class 1 'person';zimumngu, 'big white person' < murungu, class 1 'white person';zimuti, 'big tree' < muti, class 3 'tree'; (?) zinvizi, 'big river' < rwizi,class 11 'river'.

Superimposition may provide an alternative to prefix substitution,and it would appear that the young generation prefer it:

zimoto rather than zoto, 'big fire' < moto, class 3 'fire';zimwana rather than zana, 'naughty child' < mwana, class 1 'child'.

Zi- as secondary prefix is noticeable with the few vowel-commencingclass 5 nouns which already have zi- as their primary prefix:

ziziso, 'big eye' < ziso, 'eye'; zizino, 'big tooth' < zino, 'tooth'.It applies also to class 5 monosyllabic stems exhibiting (in my view)

i- as a prefix:ziibvi (also zibvi), 'big knee'; < ibui, (stem *-bvi)70 class 5 'knee'.

4.2.2 With non-syllabic prefixesOf concern here are class 9 and class 5 nouns.

(i) Class 9:zinzeve, 'big ear' < nzeve, class 9 'ear'; zitafura, 'big table' < tafura,class 9 'table'.In the following example, prefix addition tends to be preferred to

prefix substitution: zimhuka rather than buka, 'big beast' < mhuka,class 9 'animal', given by Fortune.71

70 The possible merger of the two [i] in sequence is another instance of the optionalphonological rule posited earlier allowing for the deletion of [i], when in contact with a vowel,as suggested above.

71 Fortune, An Analytical Grammar ofShona, 77.

M. LAFON 73

(ii) Class 5:Prefix addition affects class 5 nouns, irrespective of their being

primary or secondary class:a. Primary:zibadza, 'big hoe' < badza, 'hoe'; zigore, 'huge cloud' < gore, 'cloud'.

b. Secondary:Prefix addition in this case entails augmentatives of second degree,

which are often considered equivalent to the first degree, and the resultof stylistic or individual variations:

zigadzi, 'large girl' < gadzi, class 572 < mukadzi, class 1 'girl';zigomana, 'big boy' < gomana, class 573 < mukomana, class 1 'boy';zibini, 'large handle' < bini, class 5 'large handle' < mupini, class 3'handle';zidzoro, 'big head' < dzoro, class 5 'big head' < musoro, class 3'head'.Alternatively, prefix addition may occur when the first degree is,

with unvoiceable stem-initials, ill- or even un-attested:zirume, 'undesirable man' < (?) rume, class 574 < murume, class 1

'man'.In the case of non-class 5 primary nouns amenable to voicing, the

double shape (voicing and zi-) of class 5 nominal prefix associates with thetwofold augmentative construction — substitution or superimposition ofprefix — to generate a variety of forms:

mukadzi >75 gadzi, zigadzi and zimukadzi;musikana > (?) sikana, dzikana and zisikana, zidzikana, and

zimusikana even though it should be noted that the superimposition ofzi- whenever voicing of stem-initial is possible sounds very awkward, as inzimukadzi, zimusikana.

4.2.3 PluralThe formation of the class 6 plural follows the various possibilitiesreported in the marking of the secondary class in the singular.

(i) With voicing of initialThe voiced stem-initial may be conserved or sometimes, on

apparently dialectal lines, there may be a reversion to the unvoiced stem-initial which constitutes the basis of the stem:

72 Ibid.73 Ibid.74 Ibid.75 > implies derivation from the form t o t h e left of it.

74 SHONA CLASS 5 REVISITED

mhino, 'nose' > bino, 'big nose' > mabino or mapino, 'big noses';mukadzi, 'woman' > gadzi, 'big woman' > magadzi or makadzi, 'bigwomen'.

(ii) With unchanged initialThe same phenomenon may be observed in cases where the stem-

initial remains unchanged, not being amenable to voicing: rume, 'big man'> marume, 'big men'.

In both cases, however, zi- is often introduced in the plural and thefollowing forms, which coincide with those obtained as plural of class 5with superimposed zi-, are more common than the previous ones (seebelow): mazipino, 'big noses'; mazikadzi, 'big women'; mazirume, 'bigmen'.

(iii) With use of zi-When zi- is resorted to in class 5, it is retained in class 6; a behaviour

reminiscent of that of the limited set of nouns where zi- is primary, viz,ziso, 'eyes' > maziso, 'eyes'; zana, 'big child' > mazana, 'big children'.

In cases of superimposition, prefixes other than those in class 5revert to their regular plural class, which leads to a row of up to threeelicited prefixes:

zimunhu, 'big man' > mazivanhu, 'big men': (ma class 6; zi- class 5 andva- class 2);zimuti, 'big tree' > mazimiti, 'big trees': (ma- class 6; zi- class 5 andmi- class 4); andzimhuka, 'large animal' > mazimhuka, 'large animals': (ma- class 6;zi- class 5 and N- class 10).76

Accordingly, initials undergoing the class 5 voicing process yield tothe unvoiced stem-initials: zibadza, 'large hoe' > mazipadza, 'large hoes':(ma- class 6 and zi- class 5).

This rule applies when zi- acts as a primary prefix as well, withvowel-commencing stems:

ziso, 'eye' > ziziso, 'big eye' > maziziso, 'big eyes': (ma- class 6; zi-class 5 and zi- class 5).77

As mentioned earlier, this process tends to become more generalizedand is by far the commonest when, due to the nature of the stem-initial,two (or more) class 5 forms are available:

76 As the nominal prefix of c lasses 9 and 10 is identical it may be assumed, analogically, thatthe nasal in zimhuka s tands for c lass 9 nominal prefix and in mazimhuka, for that in c lass 10.

77 T h e s e forms are indicative of t h e proces s of integration of zi- into the s t em.

M. LAFON 75

gadzi and zigadzi, 'big woman' > mazikadzi, 'big women' rather thanmagadzi or makadzi;(?)sikana, dzikana, zisikana and zidzikana, 'big girl' > mazisikana,'big girls' rather than masikana or madzikana;(?) rarne and zirume, 'big man' > mazirume, 'big men' rather thanmarume.

4.2.4 AdjectivesAdjectives, whose class agreement is determined by their governing nouns,are usually not liable to connotation through a change of class. In class 5most items undergo voicing and, since they are all consonant-commencing(see Dale's table of adjectives in Shona Companion), this is sufficient todisqualify zi- as a relevant allomorph.

However, some speakers accept, and a few want superimposition ofzi- prefix following a noun in the augmentative, for some adjectivalstems at least. This can be interpreted either as an analogical extensionof the nominal prefix allomorph to the adjective, which again supportstheir sharing the same set of concords, or as a meaningful device:

zinyoka zidemadema, 'a very large jet-black snake';78

zimurungu ziguru, 'a very big white person'.79

The data produced illustrate the intricate relationship between thetwo main allomorphs of class 5 nominal prefix; namely voicing and zi-,when both its primary and secondary functions are taken into account.It follows that all augmentatives would best be considered as belongingto class 5, since they resort to the same prefix allomorphs as 'normal'nouns. Apart from phonological constraints linked to the nature of thestem-initial, the distribution of these allomorphs appears to be conditionedby the primary prefix of the noun.

This comprehensive view of class 5 leads one to regard the class 5nominal prefix as a scatter of conditioned allomorphs, both onphonological and lexical lines.

Before turning to the representation of the class 5 nominal prefix, wemust highlight the significant evidence produced by the copulativeconstruction in support of our view of /- as an allomorph of the nominalprefix, since it contributes to the construing of a 'cover form'.

78 Dale, Shona Companion, 28.79 Mkanganwi, 'An outline of the morphology of substantives . . . ' , 115.

76 SHONA CLASS 5 REVISITED

5 THE COPULATIVE EVIDENCE

Formation of the positive copulative construction in Shona resorts tothree devices, according to the nature of the noun class prefix. I reviewwhat is relevant to this discussion, leaving aside nouns included in so-called classes la, 2a and 2b, whose copulatives are specific in incorporatinga non-prefix element, ndi-.

5.1 Syllabic prefixesFor all nominals endowed with syllabic prefixes, be they noun or adjective,all copulatives except for classes la, 2a and 2b, are expressed by raisingthe tone of the otherwise low prefix.80

The well-known mnemonic rhyme furnishes ready examples of thisprocess: murume murume, vasikana vdsikana, mapanga mapanga, 'a manis a man, girls are girls, knives are knives'.

With adjectives: muti uyu murefu, class 3 'this tree is tall'.

Nominals with the syllabic allomorphs of classes 5, 9 and 10, amongwhich we tentatively include i-, behave likewise (see below for variants).

(i) Class 5: i'-; ri-\ zi-Examples pertain here to monosyllabic stems and augmentativeswith zi-\izwi izwi, 'a word is a word'; ibwe ibwe, 'a stone is a stone';rifa, 'it is inheritance' (suggested during the University of Zimbabweseminar);zigomana zigomana, 'a big boy is a big boy';zimuti ztmuti, 'a big tree is a big tree';(?) iri ziso, 'this is an eye'.81

This process affects adjectives as well: banga idzva, 'the knife isnew'.

(ii) Class 9 and class 10 i-; (dzi)Examples in these categories are limited to monosyllabic stems.

Class 9 and class 10 nouns are generally undifferentiated as far asnumber is concerned:

imbwa fmbwa, class 9 or class 10 'a dog is a dog', or 'dogs are dogs';inda inda, class 9 or class 10 'a louse is a louse' or 'lice are lice'.Adjectives show no specificity and behave as the nouns referred to

above: tafura idzi itsva, class 10 'these tables are new'.

80 The similarity exhibited by n o u n s and adject ives in the copulat ive construct ion isfurther ev idence of their c l o s e grammatical relationship.

81 For s o m e reason the template *ziso ziso was rejected outright.

M. LAFON 77

Number differentiation through nominal prefix coincides with the fewwords that form the basis for the dzi- allomorph of class 10 nominal prefix.These are limited to the aforementioned words imba/dzimba 'house'/'houses', and dzimbo, 'songs'; whose plurals allow use of the same device:

imba imba, class 9 'a house is a house'; dzimba dzimba, 'houses arehouses'; dzimbo dzimbo, 'songs are songs'.

5.2 Non-syllabic prefixesApart from a few locative examples (below) and, as already stated,nouns allocated to class la, 2a and 2b, those nouns endowed with non-syllabic prefixes are said to resort to a copulative formative consisting ofa high tone i- vowel preceding the noun:

bere ibere, class 5 'a hyena is a hyena';shumba ishumba, classes 9 and 10 'a lion is a lion' and 'lions are lions'.

Furthermore, this same device is recorded alternatively with thesyllabic allomorphs of class 5 and class 10, particularly when class 5 zi-is primary, as well as with the two words exhibiting dzi- in class 10; theresults being more colloquial:

ziso tziso, class 5 'an eye is an eye'; zino izino, 'a tooth is a tooth';dzimba idzimba, class 10 'houses are houses'; dzimbo idzimbo,'songs are songs'.However, with class 5 zi- in augmentative meaning, this possibilityseems questionable, to say the least:(?)fri, izigomana, 'this one is a big boy'.

These alternative forms thus confirm the degrammaticalization ofboth prefixes in primary usage, which zi-'s possible retention in theplural formation and dzi-'s extremely limited distribution, alsodemonstrate. As the initial syllable is no longer felt to be functioning asa prefix, it seems safe to assume that those words are assimilated toclass 5 or class 10 nouns endowed with a non-syllabic prefix.

Possible counter-examples regarding the copulative constructionare mentioned in the traditional accounts under review, but are leftunexplained. Fortune82 states that with some nouns controlling class 17agreement but bearing no overt mark of any existing syllabic prefix, be itclass 17 (feu-) or any other, the copulative is indicated by tone-raising ofthe first syllable (which entails a restructuring of the tone pattern of theword):

82 Fortune, Shona Grammatical Constructions, Vol. 1, 139.

78 SHONA CLASS 5 REVISITED

mberi, 'in front' > mbiri uko, 'it is in front there' (not *imberi uko)zasi, 'down' > z&si uko, 'it is down there' (not *izasi uko)shure, 'behind' > shure uko, 'it is behind' (not *ishure uko~)

These facts support my interpretation of 'formative' li-l as anallomorph of classes 5, 9 and 10 nominal prefix, and thus suggest theexistence of a long form of the prefix including [/'-]. This long form isretained due to phonological constraints with otherwise monosyllabicstems in most dialects and in the copulative which requires a syllable tobear the high tone.83

This interpretation also explains the absence of a segmentalembodiment of the copulative in the three locative examples givenabove, in spite of the lack of a syllabic prefix. As i- is not a class 17prefix and ndi-, which are semantically restricted, would be out ofplace, the only remaining option to mark this difference was, in myview, the raising of the tone on the first syllable.

This understanding of the copulative simplifies the treatment of themonosyllabic stems of classes 5, 9 and 10, which otherwise wouldrequire special mention of the substitution of a formative i- to theepenthetic in the copulative construction.84

Secondly, it consolidates the description of the copulativeconstruction as a whole, since all nouns except those of so-called classesla, 2a, 2b and the three aforementioned locative cases, then becomeamenable to the same procedure, viz, tone-raising of the nominal prefix.This treatment conforms to the results of Pongweni's supra-segmentalanalysis of the copulative, which concludes that the tonal behaviour ofnouns in the copulative is 'unrelated' to the phonic nature of the classnominal prefix (that is, syllabic versus non-syllabic). The dividing lineseparates nouns allocated to classes la, 2a and 2b requiring ndi-,irrespective of their prefix, from all others. This view is expressed byPongweni when he states:

It does not seem to matter whether or not a noun has a noun class prefix [that is, asyllabic or non-syllabic allomorph]... The nouns of the traditional classes la and2a, whether prefixed or unprefixed,... behave differently.85

83 This strategy is reminiscent of Lamba copulative, according to Doke's treatment ofShona c lass 8. This c las s c o r r e s p o n d s t o Common Bantu 5, the nominal prefix ofwhich is It'/i-l with vowel - or nasa l - commenc ing s t e m s , and Ir-I wi th c o n s o n a n t -commencing s t e m s . The copulative is formed by shortening the nominal prefix VCV to CV• tone raising, which causes the allomorph I i:-l to revert to the full form of the prefix, I i/i-l,which is then reduced to CV in C. Doke, Textbook of Lamba Grammar (Johannesburg,Witwatersrand University Press, 1938), 309.

84 Fortune, An Analytical Grammar of Shona, 371.85 P o n g w e n i , Studies in Shona Phonetics .... 2 0 4 - 5 .

M. LAFON 79

6 REPRESENTATION OF CLASS 5 NOMINAL PREFIX AS A SCATTEROF ALLOMORPHS

If we acknowledge the various allomorphs displayed by class 5 nominalprefix we can then posit a scatter of forms whose distribution, only partiallyattributable to phonological constraints, is also related to lexical andsyntactical constraints.

l)Raw stems:i) vowel commencing: zi-ii) consonant commencing:

a) monosyllabic: i-, 0-, or ri- (dialectally)b) with stem-initials [p, t, k, ch, tsv, pf, sv]: voicing or 0- in

some adoptivesc) other consonants non subject to voicing: 0-

2) Prefixed-nouns: zi- (augmentative)3) Copulative — tone-raising of prefix:

a) with syllabic allomorphs: i-/z(-b) with non-syllabic ones and zi- in some cases:

re-introduction of i-.

Any arbitrary symbol could be construed to embody the voicingprocess since, clearly, the changes entailed are no longer caused by asegmental sequence but occur by virtue of the entry of these nouns intoclass 5, as is evidenced by the hesitant behaviour of recent adoptives.The scarcity of words resisting this conditioning is indicative of thestrength of the process at work.

Fortune expressed the view that changes in the stem-initial consonantwere due to a former *ri- prefix. Even if we accept the historicalperspective, it seems unlikely that an r consonant should have playedany part in such an important process only to disappear completely.Such changes are generally attributed, in Bantu, to the influence of aclosed vowel on a following consonant. This is especially the case inlanguages such as Shona, which have undergone a reduction from a 7-vowel system to a 5-vowel system.86

86 The classical hypothesis on Common or Proto-Bantu posits a 7-vowel system, the twoextreme front vowels being represented by cedillas under i and u, as I and u.

Historically, the inclusion of the latent vowel in the prefix suggests the followingsequence for the main allomorphs: from a nominal prefix construed as *iri- a first stage woulddelete the 'r' consonant which is in a weak position. The two Y in contact would then bemerged, entailing the modification, when possible, of the stem-initial (ie 'voicing'), and disappearin all except monosyllabic stems and copulative phrases. Hence,

1-CW monosyllabic stems*iri-C... > */i-C... > CV(CV)CW longer stems

i-C... copulative.

80 SHONA CLASS 5 REVISITED

Therefore, if we wish to coin symbols reminiscent of real features,then V- for voicing might be appropriate, with *i- or perhaps */- as a 'coverform' catering also for the copulative. If we extend the scope of V- toencompass the lack of change occurring with non-voiceable initials, anotherstatement of the scatter of allomorphs of class 5 nominal prefix couldread:

Class 5 nominal prefix */-:• /-: monosyllabic stems (dialectally); copulative construction

except some words with zi- and ri-• 0-, rarely ri-: monosyllabic stems (dialectally)• V- (voicing of voiceable initials, no change for non-voiceable

ones): consonant-commencing stems• zi-: vowel-commencing stems and augmentatives with non-

voiceable initials, including prefixes.

CONCLUSION

I have argued that the rationale behind positing *ri- as a general coverform for Shona class 5 nominal prefix should be abandoned in favour ofthe concept of a cluster or scatter of allomorphs. This cluster couldinclude I ri-1 itself, as a real if rare form 11-1, and positing IV-1 for voicing.Furthermore, this cluster could in its entirety, be subsumed by */- sincethe vowel [«-] surfaces in the majority of cases, in the copulative at least.This theory would obviate the need to resort to an epenthetic vowel in thecase of monosyllabic stems and would, in turn, allow for some simplificationin the general account of the copulative.

The sign * implies the word border. For vowel-commencing stems, *izi- would have mergedwith the initial vowel of the stem, leading to a restructured 'i-zv..., which would then evolveinto zv... and i-zv... for the copulative (V representing the result of vowel coalescence, i + ...):*tzi-VCV(CV) > *i-2uCV(CV) > zvCV(CV),

t-zvCV(CV) or ziS CV(CV) In the copulative.In the case of superimposition of the prefix, the scope of zi- would be extended, the possiblelong copulative izi- then being viewed either as a retention of a longer, prior form, or as ananalogically supplemented device.